Files
MyFSIO/app/gc.py

597 lines
22 KiB
Python

from __future__ import annotations
import json
import logging
import os
import shutil
import threading
import time
from dataclasses import dataclass, field
from datetime import datetime, timezone
from pathlib import Path
from typing import Any, Dict, List, Optional
logger = logging.getLogger(__name__)
@dataclass
class GCResult:
temp_files_deleted: int = 0
temp_bytes_freed: int = 0
multipart_uploads_deleted: int = 0
multipart_bytes_freed: int = 0
lock_files_deleted: int = 0
orphaned_metadata_deleted: int = 0
orphaned_versions_deleted: int = 0
orphaned_version_bytes_freed: int = 0
empty_dirs_removed: int = 0
errors: List[str] = field(default_factory=list)
execution_time_seconds: float = 0.0
def to_dict(self) -> dict:
return {
"temp_files_deleted": self.temp_files_deleted,
"temp_bytes_freed": self.temp_bytes_freed,
"multipart_uploads_deleted": self.multipart_uploads_deleted,
"multipart_bytes_freed": self.multipart_bytes_freed,
"lock_files_deleted": self.lock_files_deleted,
"orphaned_metadata_deleted": self.orphaned_metadata_deleted,
"orphaned_versions_deleted": self.orphaned_versions_deleted,
"orphaned_version_bytes_freed": self.orphaned_version_bytes_freed,
"empty_dirs_removed": self.empty_dirs_removed,
"errors": self.errors,
"execution_time_seconds": self.execution_time_seconds,
}
@property
def total_bytes_freed(self) -> int:
return self.temp_bytes_freed + self.multipart_bytes_freed + self.orphaned_version_bytes_freed
@property
def has_work(self) -> bool:
return (
self.temp_files_deleted > 0
or self.multipart_uploads_deleted > 0
or self.lock_files_deleted > 0
or self.orphaned_metadata_deleted > 0
or self.orphaned_versions_deleted > 0
or self.empty_dirs_removed > 0
)
@dataclass
class GCExecutionRecord:
timestamp: float
result: dict
dry_run: bool
def to_dict(self) -> dict:
return {
"timestamp": self.timestamp,
"result": self.result,
"dry_run": self.dry_run,
}
@classmethod
def from_dict(cls, data: dict) -> GCExecutionRecord:
return cls(
timestamp=data["timestamp"],
result=data["result"],
dry_run=data.get("dry_run", False),
)
class GCHistoryStore:
def __init__(self, storage_root: Path, max_records: int = 50) -> None:
self.storage_root = storage_root
self.max_records = max_records
self._lock = threading.Lock()
def _get_path(self) -> Path:
return self.storage_root / ".myfsio.sys" / "config" / "gc_history.json"
def load(self) -> List[GCExecutionRecord]:
path = self._get_path()
if not path.exists():
return []
try:
with open(path, "r", encoding="utf-8") as f:
data = json.load(f)
return [GCExecutionRecord.from_dict(d) for d in data.get("executions", [])]
except (OSError, ValueError, KeyError) as e:
logger.error("Failed to load GC history: %s", e)
return []
def save(self, records: List[GCExecutionRecord]) -> None:
path = self._get_path()
path.parent.mkdir(parents=True, exist_ok=True)
data = {"executions": [r.to_dict() for r in records[: self.max_records]]}
try:
with open(path, "w", encoding="utf-8") as f:
json.dump(data, f, indent=2)
except OSError as e:
logger.error("Failed to save GC history: %s", e)
def add(self, record: GCExecutionRecord) -> None:
with self._lock:
records = self.load()
records.insert(0, record)
self.save(records)
def get_history(self, limit: int = 50, offset: int = 0) -> List[GCExecutionRecord]:
return self.load()[offset : offset + limit]
def _dir_size(path: Path) -> int:
total = 0
try:
for f in path.rglob("*"):
if f.is_file():
try:
total += f.stat().st_size
except OSError:
pass
except OSError:
pass
return total
def _file_age_hours(path: Path) -> float:
try:
mtime = path.stat().st_mtime
return (time.time() - mtime) / 3600.0
except OSError:
return 0.0
class GarbageCollector:
SYSTEM_ROOT = ".myfsio.sys"
SYSTEM_TMP_DIR = "tmp"
SYSTEM_MULTIPART_DIR = "multipart"
SYSTEM_BUCKETS_DIR = "buckets"
BUCKET_META_DIR = "meta"
BUCKET_VERSIONS_DIR = "versions"
INTERNAL_FOLDERS = {".meta", ".versions", ".multipart"}
def __init__(
self,
storage_root: Path,
interval_hours: float = 6.0,
temp_file_max_age_hours: float = 24.0,
multipart_max_age_days: int = 7,
lock_file_max_age_hours: float = 1.0,
dry_run: bool = False,
max_history: int = 50,
io_throttle_ms: int = 10,
) -> None:
self.storage_root = Path(storage_root)
self.interval_seconds = interval_hours * 3600.0
self.temp_file_max_age_hours = temp_file_max_age_hours
self.multipart_max_age_days = multipart_max_age_days
self.lock_file_max_age_hours = lock_file_max_age_hours
self.dry_run = dry_run
self._timer: Optional[threading.Timer] = None
self._shutdown = False
self._lock = threading.Lock()
self._scanning = False
self._scan_start_time: Optional[float] = None
self._io_throttle = max(0, io_throttle_ms) / 1000.0
self.history_store = GCHistoryStore(storage_root, max_records=max_history)
def start(self) -> None:
if self._timer is not None:
return
self._shutdown = False
self._schedule_next()
logger.info(
"GC started: interval=%.1fh, temp_max_age=%.1fh, multipart_max_age=%dd, lock_max_age=%.1fh, dry_run=%s",
self.interval_seconds / 3600.0,
self.temp_file_max_age_hours,
self.multipart_max_age_days,
self.lock_file_max_age_hours,
self.dry_run,
)
def stop(self) -> None:
self._shutdown = True
if self._timer:
self._timer.cancel()
self._timer = None
logger.info("GC stopped")
def _schedule_next(self) -> None:
if self._shutdown:
return
self._timer = threading.Timer(self.interval_seconds, self._run_cycle)
self._timer.daemon = True
self._timer.start()
def _run_cycle(self) -> None:
if self._shutdown:
return
try:
self.run_now()
except Exception as e:
logger.error("GC cycle failed: %s", e)
finally:
self._schedule_next()
def run_now(self, dry_run: Optional[bool] = None) -> GCResult:
if not self._lock.acquire(blocking=False):
raise RuntimeError("GC is already in progress")
effective_dry_run = dry_run if dry_run is not None else self.dry_run
try:
self._scanning = True
self._scan_start_time = time.time()
start = self._scan_start_time
result = GCResult()
original_dry_run = self.dry_run
self.dry_run = effective_dry_run
try:
self._clean_temp_files(result)
self._clean_orphaned_multipart(result)
self._clean_stale_locks(result)
self._clean_orphaned_metadata(result)
self._clean_orphaned_versions(result)
self._clean_empty_dirs(result)
finally:
self.dry_run = original_dry_run
result.execution_time_seconds = time.time() - start
if result.has_work or result.errors:
logger.info(
"GC completed in %.2fs: temp=%d (%.1f MB), multipart=%d (%.1f MB), "
"locks=%d, meta=%d, versions=%d (%.1f MB), dirs=%d, errors=%d%s",
result.execution_time_seconds,
result.temp_files_deleted,
result.temp_bytes_freed / (1024 * 1024),
result.multipart_uploads_deleted,
result.multipart_bytes_freed / (1024 * 1024),
result.lock_files_deleted,
result.orphaned_metadata_deleted,
result.orphaned_versions_deleted,
result.orphaned_version_bytes_freed / (1024 * 1024),
result.empty_dirs_removed,
len(result.errors),
" (dry run)" if effective_dry_run else "",
)
record = GCExecutionRecord(
timestamp=time.time(),
result=result.to_dict(),
dry_run=effective_dry_run,
)
self.history_store.add(record)
return result
finally:
self._scanning = False
self._scan_start_time = None
self._lock.release()
def run_async(self, dry_run: Optional[bool] = None) -> bool:
if self._scanning:
return False
t = threading.Thread(target=self.run_now, args=(dry_run,), daemon=True)
t.start()
return True
def _system_path(self) -> Path:
return self.storage_root / self.SYSTEM_ROOT
def _throttle(self) -> bool:
if self._shutdown:
return True
if self._io_throttle > 0:
time.sleep(self._io_throttle)
return self._shutdown
def _list_bucket_names(self) -> List[str]:
names = []
try:
for entry in self.storage_root.iterdir():
if entry.is_dir() and entry.name != self.SYSTEM_ROOT:
names.append(entry.name)
except OSError:
pass
return names
def _clean_temp_files(self, result: GCResult) -> None:
tmp_dir = self._system_path() / self.SYSTEM_TMP_DIR
if not tmp_dir.exists():
return
try:
for entry in tmp_dir.iterdir():
if self._throttle():
return
if not entry.is_file():
continue
age = _file_age_hours(entry)
if age < self.temp_file_max_age_hours:
continue
try:
size = entry.stat().st_size
if not self.dry_run:
entry.unlink()
result.temp_files_deleted += 1
result.temp_bytes_freed += size
except OSError as e:
result.errors.append(f"temp file {entry.name}: {e}")
except OSError as e:
result.errors.append(f"scan tmp dir: {e}")
def _clean_orphaned_multipart(self, result: GCResult) -> None:
cutoff_hours = self.multipart_max_age_days * 24.0
bucket_names = self._list_bucket_names()
for bucket_name in bucket_names:
if self._shutdown:
return
for multipart_root in (
self._system_path() / self.SYSTEM_MULTIPART_DIR / bucket_name,
self.storage_root / bucket_name / ".multipart",
):
if not multipart_root.exists():
continue
try:
for upload_dir in multipart_root.iterdir():
if self._throttle():
return
if not upload_dir.is_dir():
continue
self._maybe_clean_upload(upload_dir, cutoff_hours, result)
except OSError as e:
result.errors.append(f"scan multipart {bucket_name}: {e}")
def _maybe_clean_upload(self, upload_dir: Path, cutoff_hours: float, result: GCResult) -> None:
manifest_path = upload_dir / "manifest.json"
age = _file_age_hours(manifest_path) if manifest_path.exists() else _file_age_hours(upload_dir)
if age < cutoff_hours:
return
dir_bytes = _dir_size(upload_dir)
try:
if not self.dry_run:
shutil.rmtree(upload_dir, ignore_errors=True)
result.multipart_uploads_deleted += 1
result.multipart_bytes_freed += dir_bytes
except OSError as e:
result.errors.append(f"multipart {upload_dir.name}: {e}")
def _clean_stale_locks(self, result: GCResult) -> None:
buckets_root = self._system_path() / self.SYSTEM_BUCKETS_DIR
if not buckets_root.exists():
return
try:
for bucket_dir in buckets_root.iterdir():
if self._shutdown:
return
if not bucket_dir.is_dir():
continue
locks_dir = bucket_dir / "locks"
if not locks_dir.exists():
continue
try:
for lock_file in locks_dir.iterdir():
if self._throttle():
return
if not lock_file.is_file() or not lock_file.name.endswith(".lock"):
continue
age = _file_age_hours(lock_file)
if age < self.lock_file_max_age_hours:
continue
try:
if not self.dry_run:
lock_file.unlink(missing_ok=True)
result.lock_files_deleted += 1
except OSError as e:
result.errors.append(f"lock {lock_file.name}: {e}")
except OSError as e:
result.errors.append(f"scan locks {bucket_dir.name}: {e}")
except OSError as e:
result.errors.append(f"scan buckets for locks: {e}")
def _clean_orphaned_metadata(self, result: GCResult) -> None:
bucket_names = self._list_bucket_names()
for bucket_name in bucket_names:
if self._shutdown:
return
legacy_meta = self.storage_root / bucket_name / ".meta"
if legacy_meta.exists():
self._clean_legacy_metadata(bucket_name, legacy_meta, result)
new_meta = self._system_path() / self.SYSTEM_BUCKETS_DIR / bucket_name / self.BUCKET_META_DIR
if new_meta.exists():
self._clean_index_metadata(bucket_name, new_meta, result)
def _clean_legacy_metadata(self, bucket_name: str, meta_root: Path, result: GCResult) -> None:
bucket_path = self.storage_root / bucket_name
try:
for meta_file in meta_root.rglob("*.meta.json"):
if self._throttle():
return
if not meta_file.is_file():
continue
try:
rel = meta_file.relative_to(meta_root)
object_key = rel.as_posix().removesuffix(".meta.json")
object_path = bucket_path / object_key
if not object_path.exists():
if not self.dry_run:
meta_file.unlink(missing_ok=True)
result.orphaned_metadata_deleted += 1
except (OSError, ValueError) as e:
result.errors.append(f"legacy meta {bucket_name}/{meta_file.name}: {e}")
except OSError as e:
result.errors.append(f"scan legacy meta {bucket_name}: {e}")
def _clean_index_metadata(self, bucket_name: str, meta_root: Path, result: GCResult) -> None:
bucket_path = self.storage_root / bucket_name
try:
for index_file in meta_root.rglob("_index.json"):
if self._throttle():
return
if not index_file.is_file():
continue
try:
with open(index_file, "r", encoding="utf-8") as f:
index_data = json.load(f)
except (OSError, json.JSONDecodeError):
continue
keys_to_remove = []
for key in index_data:
rel_dir = index_file.parent.relative_to(meta_root)
if rel_dir == Path("."):
full_key = key
else:
full_key = rel_dir.as_posix() + "/" + key
object_path = bucket_path / full_key
if not object_path.exists():
keys_to_remove.append(key)
if keys_to_remove:
if not self.dry_run:
for k in keys_to_remove:
index_data.pop(k, None)
if index_data:
try:
with open(index_file, "w", encoding="utf-8") as f:
json.dump(index_data, f)
except OSError as e:
result.errors.append(f"write index {bucket_name}: {e}")
continue
else:
try:
index_file.unlink(missing_ok=True)
except OSError:
pass
result.orphaned_metadata_deleted += len(keys_to_remove)
except OSError as e:
result.errors.append(f"scan index meta {bucket_name}: {e}")
def _clean_orphaned_versions(self, result: GCResult) -> None:
bucket_names = self._list_bucket_names()
for bucket_name in bucket_names:
if self._shutdown:
return
bucket_path = self.storage_root / bucket_name
for versions_root in (
self._system_path() / self.SYSTEM_BUCKETS_DIR / bucket_name / self.BUCKET_VERSIONS_DIR,
self.storage_root / bucket_name / ".versions",
):
if not versions_root.exists():
continue
try:
for key_dir in versions_root.iterdir():
if self._throttle():
return
if not key_dir.is_dir():
continue
self._clean_versions_for_key(bucket_path, versions_root, key_dir, result)
except OSError as e:
result.errors.append(f"scan versions {bucket_name}: {e}")
def _clean_versions_for_key(
self, bucket_path: Path, versions_root: Path, key_dir: Path, result: GCResult
) -> None:
try:
rel = key_dir.relative_to(versions_root)
except ValueError:
return
object_path = bucket_path / rel
if object_path.exists():
return
version_files = list(key_dir.glob("*.bin")) + list(key_dir.glob("*.json"))
if not version_files:
return
for vf in version_files:
try:
size = vf.stat().st_size if vf.suffix == ".bin" else 0
if not self.dry_run:
vf.unlink(missing_ok=True)
if vf.suffix == ".bin":
result.orphaned_version_bytes_freed += size
result.orphaned_versions_deleted += 1
except OSError as e:
result.errors.append(f"version file {vf.name}: {e}")
def _clean_empty_dirs(self, result: GCResult) -> None:
targets = [
self._system_path() / self.SYSTEM_TMP_DIR,
self._system_path() / self.SYSTEM_MULTIPART_DIR,
self._system_path() / self.SYSTEM_BUCKETS_DIR,
]
for bucket_name in self._list_bucket_names():
targets.append(self.storage_root / bucket_name / ".meta")
targets.append(self.storage_root / bucket_name / ".versions")
targets.append(self.storage_root / bucket_name / ".multipart")
for root in targets:
if not root.exists():
continue
self._remove_empty_dirs_recursive(root, root, result)
def _remove_empty_dirs_recursive(self, path: Path, stop_at: Path, result: GCResult) -> bool:
if self._shutdown:
return False
if not path.is_dir():
return False
try:
children = list(path.iterdir())
except OSError:
return False
all_empty = True
for child in children:
if self._throttle():
return False
if child.is_dir():
if not self._remove_empty_dirs_recursive(child, stop_at, result):
all_empty = False
else:
all_empty = False
if all_empty and path != stop_at:
try:
if not self.dry_run:
path.rmdir()
result.empty_dirs_removed += 1
return True
except OSError:
return False
return all_empty
def get_history(self, limit: int = 50, offset: int = 0) -> List[dict]:
records = self.history_store.get_history(limit, offset)
return [r.to_dict() for r in records]
def get_status(self) -> dict:
status: Dict[str, Any] = {
"enabled": not self._shutdown or self._timer is not None,
"running": self._timer is not None and not self._shutdown,
"scanning": self._scanning,
"interval_hours": self.interval_seconds / 3600.0,
"temp_file_max_age_hours": self.temp_file_max_age_hours,
"multipart_max_age_days": self.multipart_max_age_days,
"lock_file_max_age_hours": self.lock_file_max_age_hours,
"dry_run": self.dry_run,
"io_throttle_ms": round(self._io_throttle * 1000),
}
if self._scanning and self._scan_start_time:
status["scan_elapsed_seconds"] = time.time() - self._scan_start_time
return status